The 2% gains that 10x your inference | Lessons from AWS on optimizing VLMs @aicouncilconf
The 2% gains that 10x your inference | Lessons from AWS on optimizing VLMs  @aicouncilconf
Uploaded April 2026 | Updated November 2026, 1 day ago
Vik Korrapati, co-founder & CTO of Moondream, shares how his distributed systems background at AWS has shaped Moondream's engineering-first approach to optimizing visual language model inference.

Key takeaways:

- Pursuing incremental improvements (1-3%) compounds to 10x systems
- Moondream built a custom inference engine tailored for VLM performance characteristics
- Optimizing end-to-end, from custom kernels to image decoding libraries
- Relentlessly tackling the bottleneck, wherever it moves


Read the Q&A: substack.com/home/post/p-193832852
👉 Sign up for our "No BS" Newsletter to get the latest technical data & AI content: aicouncil.com/newsletter

ABOUT AI COUNCIL:
AI Council brings together the brightest minds in data to share industry knowledge, technical architectures and best practices in building cutting edge data & AI systems and tools.

FIND US:
Twitter: twitter.com/AICouncilConf
LinkedIn: linkedin.com/company/datacouncil-ai
Website: datacouncil.ai


#Moondream #VLM #ModelInference #EngineeringExcellence #AICouncil
The 2% gains that 10x your inference | Lessons from AWS on optimizing VLMsQ&A with Scott Breitenother, Kilo: Engineers need to be the CEOs of agents. Are they ready?Agentic AI: From Risk Awareness to Practical Control | Noma SecurityShould agents be durable? | RenderGuardrails for the Future AI Safety and Responsible AI in PracticeAI Launchpad 2025: MooncakeThe Middle Ground: Balancing Batch and Real-Time Processing in a Data LakehouseAI Launchpad 2025: TopKRunning Millions of (Millisecond) AI Sandboxes without Breaking the Piggy Bank | UnikraftMove Fast and Dont Break Things   How to Build a Data Platform that Scales with your OrganizationBenchmarking AI Agents Against Realistic Analytical Tasks with ADE-benchRevolutionize AI Engineering with AutoGen
AI Council |

The 2% gains that 10x your inference | Lessons from AWS on optimizing VLMs

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER